Merge remote-tracking branch 'nccl/master' into develop

[ROCm/rccl commit: a6bf9bfc9e]
2025-04-23 20:46:36 -07:00
parent 07620c7efd 8f0723e904
commit d045d0ca23
118 changed files with 12789 additions and 3358 deletions
@@ -223,7 +223,9 @@ namespace RcclUnitTesting
        NCCLCHECK(ncclGroupStart());
        for (int rank = 0; rank < numRanks; rank++)
          NCCLCHECK(ncclAllReduce(gpuInput[rank], gpuOutput[rank], N, ncclInt, ncclSum, comms[rank], stream[rank]));
-        NCCLCHECK(ncclGroupEnd());
+        ncclResult_t res = ncclGroupEnd();
+
+        if (res != ncclSuccess) continue;

        const auto start = Clock::now();